Actual, Adjacent, and Potential Competition
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Journal of Transport Economics and Policy, Volume 35, Part 2, May 2001, pp.239± 256
Actual, Adjacent, and Potential Competition
Estimating the Full EŒect of Southwest Airlines
Steven A. Morrison
Address for correspondence: Department of Economics, Northeastern University, Boston,
MA 02115 (sam1@neu.edu). This paper is based on research undertaken on behalf of
Southwest Airlines. The views expressed are the author’s own and are not necessarily those
of Southwest Airlines. The author would like to thank Tae Oum, CliŒWinston, and an
anonymous referee for helpful comments.
Abstract
Southwest Airlines is frequently credited with having an important influence on the success
of airline deregulation in the United States. This paper uses an original set of competition
variables to estimate the extent of that influence in 1998. The estimated savings Ð due to
actual, adjacent, and potential competition from Southwest Ð were $12.9 billion.
Southwest’s low fares were directly responsible for $3.4 billion of these savings to
passengers. The remaining $9.5 billion represents the effect that actual, adjacent, and
potential competition from Southwest had on other carriers’ fares. These savings amount
to 20 per cent of the airline industry’s 1998 domestic scheduled passenger revenue and
slightly more than half the fare reductions attributed to airline deregulation. From a policy
perspective, these results are both troubling and encouraging. On the one hand, it is
troubling to find that a large part of the fare reductions from airline deregulation is due to
one carrier. On the other hand, if entry by a carrier with the appropriate characteristics can
make such a difference, policies that encourage entry Ð for example, relaxing the
restriction on entry by foreign-owned carriers Ð may have a large impact on passenger
welfare.
Date of receipt of ® nal manuscript: July 2000
239Journal of Transport Economics and Policy Volume 35, Part 2
Introduction
``Probably the most signi® cant development in the U.S. airline industry
during the past decade has been the continued expansion of Southwest Air-
lines and the resurgence of low-fare entry generally.’’
Transportation Research Board (1999)
Southwest Airlines has a unique position in the US domestic aviation
industry. It was one of two carriers (along with Paci® c Southwest Airlines)
whose low fares as intrastate carriers not regulated by the federal gov-
ernment led to airline deregulation in 1978.1 Indeed, its combination of
low fares, high frequency, and point-to-point service has become a model
for other carriers both in the United States and abroad. As exempli® ed by
the quote above, Southwest Airlines is often cited as a major reason for the
success of airline deregulation in the United States. This paper uses an
original set of competition variables to quantify the impact that Southwest
Airlines has on airfares through actual, adjacent, and potential competi-
tion.
Previous Research
Previous research has estimated Southwest’s aggregate impact on fares on
the routes that it serves.2 Dresner et al. (1996) hypothesised that the eŒect
of Southwest (and other low-fare carriers) may be greater than previously
estimated because of possible spillover eŒects that Southwest’s service on
one route has on adjacent competitive routes that involve nearby airports.
They found a signi® cant eŒect of service on adjacent competitive routes
but did not aggregate the results to estimate the aggregate impact of
Southwest. This paper’s contribution is to specify the eŒect that Southwest
may have on a route’ s fares in an original disaggregate way, which
includes the eŒect of various forms of actual, adjacent, and potential
competition from Southwest. The results are then aggregated to obtain an
estimate of the eŒect that Southwest has had on airfares. This estimate is
1
See Breyer (1982). Southwest Airlines started service in 1971 as a Texas intrastate carrier
serving three Texas cities. It began interstate service shortly after airlines were deregulated
in 1978. In 1999, it served 57 airports (in 29 of the 48 contiguous states) and had operating
revenue of $4.7 billion, which placed it seventh among US passenger airlines.
2
See, for example, Bennett and Craun (1993) and US Department of Transportation
(1996).
240Actual, Adjacent, and Potential Competition Morrison
compared with estimates of the fare reductions attributed to airline
deregulation to gauge Southwest’s role in this policy’ s success.
Methodology
Airfares on a route can be in¯ uenced by an airline in three ways. First, the
airline can serve the route in question. Second, the airline can serve an
adjacent route that consumers view as a reasonable substitute for the route
in question. Third, although the airline may not serve the route in question
or an adjacent route, it may still aŒect fares on the route if airlines lower
their fares to make entry by potential competitors less attractive.3 This
could happen if the airline has a presence at the airports in question or at
nearby airports.
Figure 1 illustrates the various ways that an airline may aŒect fares on a
particular route. In the ® gure, the route of interest is from airport A to
airport B. The dashed circles around airports A and B form the bound-
aries of zones of in¯ uence Ð any carrier serving an airport within those
zones may exert a competitive in¯ uence on the carriers serving the airports
in question. In the diagram, airports C and D fall within the zones of A
and B, respectively, and airport E is outside both zones.
The possible ways in which a carrier may aŒect fares on a given route
A-B are:
. It may serve the route A-B (denoted by AAS, for At the origin airport,
At the destination airport, and Serves the route).
. It may serve a route A-D, where airport D is ``near’ ’ airport B (denoted
by ANS, for At, Near, Serves).
. It may serve a route C-D, where C is near airport A and D is near
airport B (denoted by NNS, for Near, Near, Serves).
. It may serve both airports A and B but not serve the route (denoted by
AAX, for At, At, does not serve the route).
. It may serve airport A and airport D, which is near airport B, but not
serve the route (denoted by ANX, for At, Near, does not serve the
route).
. It may serve airports C and D but not serve the route (denoted by
NNX, for Near, Near, does not serve the route).
3
The role of potential competition was formalised by Baumol et al. (1982). In their model,
the threat of entry by potential competitors disciplines those ® rms serving the market to
price at a level that maximises consumer welfare.
241Journal of Transport Economics and Policy Volume 35, Part 2
Figure 1
Ways in which a Carrier may aŒect Fares on Route A± B
. It may serve a route A-E, where E is not near B (denoted by AFS, for
At, Far, Serves).
. It may serve a route C-E, where C is near A, but E is not near B
(denoted by NFS, for Near, Far, Serves).
In the ® rst of these cases, the airline serves the route in question. In the next
two cases, the airline provides service on adjacent routes. In the remaining
® ve cases, a carrier provides various degrees of potential competition (the
242Actual, Adjacent, and Potential Competition Morrison
extent of which will be estimated) by serving one or both of the airports A
and B or airports nearby, but not serving routes connecting them.4
The Data and the Model
The primary dataset for this study was the US Department of Trans-
portation’s Ticket Origin and Destination Survey (Data Bank 1A) for the
calendar year 1998, a 10 per cent sample of airline tickets, which carriers
report quarterly to the Department.5 Although the eŒect of Southwest
Airlines will be empirically estimated below, the extent of its possible
in¯ uence is shown in Table 1, which shows the percentage of domestic
passenger miles ¯ own on routes in each of the categories above. To avoid
double counting, Southwest’ s (potential) in¯ uence on each route was
classi® ed into the highest category for which it quali® ed. During 1998,
Southwest was in a position to (potentially) in¯ uence fares on the routes it
served and on adjacent routes amounting to 44.8 per cent of domestic
passenger miles ¯ own. In addition, in its role as a potential competitor, it
was in a position to (potentially) in¯ uence fares on routes accounting for
an additional 49.4 per cent of domestic passenger miles. Thus, Southwest
was in a position to potentially in¯ uence fares on routes comprising 94 per
cent of domestic passenger miles.6
4
The last two cases are considered as potential competition even though the carrier oŒers
service on A-E or C-E because airport E is outside the zone of in¯ uence for the A-B route.
5
In order to be reasonably assured that the tickets for a particular trip re¯ ect travel from
one origin to one destination, only tickets with one (directional) destination were used.
Round trips had to return to the initial point of departure (with no ground segments, i.e.,
``open jaw’’ tickets). Only one-way tickets with two or fewer coupons (i.e., ¯ ight segments)
and round-trip tickets with two or fewer coupons on the outbound and return legs were
used. Further, only tickets involving airports in the 48 contiguous states were used.
Finally, tickets that were coded with the generic codes for Chicago, Detroit, New York
City, and Washington, DC were omitted because it was not possible to identify the airport
actually used.
6
What characterises those routes that are beyond any (potential) impact of Southwest? The
20 most heavily travelled routes (based on passenger miles) that Southwest did not
(potentially) in¯ uence in 1998 (amounting to 24 per cent of the passenger miles of that
unaŒected group) involved routes between the following 13 airports: Atlanta, BuŒalo,
Charlotte, Denver, Hartford, LaGuardia, Memphis, Minneapolis-St. Paul, Newark,
Raleigh-Durham, Richmond, Philadelphia, and Pittsburgh. Eight of these airports are
hubs for major airlines. Eleven are in the eastern US. One is slot constrained. (In 1999,
Southwest entered Hartford and Raleigh-Durham.)
243Journal of Transport Economics and Policy Volume 35, Part 2
Table 1
Competitive Pro® le of Southwest Airlines in 1998
Category Variable Percentage of
Domestic Passenger Miles
Actual Competition 44.8%
At-At-Serves 21.0%
At-Near-Serves 18.0%
Near-Near-Serves 5.7%
Potential Competition 49.4%
At-At-Does not Serve 6.5%
At-Near-Does not Serve 2.7%
Near-Near-Does not Serve 0.1%
At-Far-Serves 30.7%
Near-Far-Serves 9.4%
Total 94.2%
Source: Author’s calculations. See text for a detailed explanation of the variables. Cal-
culations are based on a zone of in¯ uence with a 75-mile radius. Totals may not add due
to rounding.
The standard approach to assessing the eŒect of competition Ð both
actual and potential Ð on airfares is to estimate a fare equation.7 In a
fare equation, the average fare on a route is regressed on measures of
actual and potential competition and other control variables that are
believed to in¯ uence the cost and demand characteristics of the route.
Here, quarterly data from the 1,000 most heavily travelled routes in 1998
were used to estimate a fare equation to assess the eŒect of Southwest
Airlines on fares.8
The Dependent Variable
For each ticket in the sample, the price each passenger paid was reduced by
the amount of ticket taxes and passenger facility charges (PFCs).9 Average
one-way fare (logarithm) was then calculated for each route in the sample.10
7
This fare equation may be regarded as a reduced form equation derived from a structural
model. See, for example, Bailey and Panzar (1981), or Morrison and Winston (1995).
8
The 1,000 routes account for 70 per cent of passengers and 67 per cent of passenger miles
in the full dataset containing more than 29,000 routes.
9
During the ® rst three quarters of 1998, the federal domestic ticket tax was 9 per cent of the
ticket price plus $1 per ¯ ight segment. During the fourth quarter of 1998, the tax was 8 per
cent plus $2 per segment. 290 airports assessed PFCs during 1998; all but one airport
charged the statutory maximum of $3 per enplanement; one airport charged $2.
10
Because of possible coding errors in the data carriers submit to the Department of
Transportation, the US General Accounting O ce’s (1990) fare screen was used to screen
out fares that seemed too high. In order to keep frequent ¯ ier tickets, a low-fare screen
was not used. The eŒect on the results of omitting frequent ¯ ier tickets is discussed below
in the section on sensitivity.
244Actual, Adjacent, and Potential Competition Morrison
The Competition Variables
As above, Southwest’s (potential) in¯ uence on each route was classi® ed
into the highest category for which it quali® ed. In addition, competition
variables giving the number of carriers from each group serving the route
were included for three other groups of carriers:11 Low-Fare carriers
(other than Southwest), Major carriers (other than Southwest), and Other
carriers.12 A zone of in¯ uence with a radius of 75 miles was used because it
provided the best ® t compared with zones of 25, 50, 100, and 125 miles.13
Control Variables
Although we are primarily interested in the eŒect that Southwest Airlines
has on fares, it is important to control for other in¯ uences in order that we
are reasonably certain that the competition variables capture just that,
rather than some other eŒect.
Distance:
Other things being equal, we expect fares to be higher the longer the trip
because costs are higher. The logarithm of the (great circle) distance from
the origin to the destination was used.
Density:
Because of ® xed station costs and economies of aircraft size, airline service
is characterised by economies of density.14 An airport is classi® ed into one
of four ``hub’’ categories (Large, Medium, Small, and Non) based on its
11
A carrier was considered as serving a route if it had at least a 10 per cent share of the
route’s passengers. Attempts to estimate an equation with a complete set of competition
variables to capture the eŒect of adjacent and potential competition for major, low fare,
and other carriers resulted in implausible estimates for all competition variables.
12
The Low-Fare carriers identi® ed by the US Department of Transportation that were
operating during 1998 (other than Southwest) were AirTran, American Trans Air,
Frontier, Kiwi, ProAir, Reno, Spirit, Tower Air, Vanguard, and Western Paci® c. The US
Department of Transportation de® nes a Major carrier as one with annual revenue over
$1 billion. In 1998, the Major (passenger) carriers (in addition to Southwest) were
Alaska, American, America West, Continental, Delta, Northwest, Trans World, United,
and US Airways. The Other carriers used in this analysis were Casino Express, Eastwind,
Hawaiian, Midway, and Midwest Express. Carriers with fewer than 1,000 sampled
passengers during 1998 were not assigned a competition variable, nor were trips
involving changing airlines, i.e., interline connections.
13
Dresner et al. (1996) used a 50-mile radius to de® ne adjacent competitive routes.
Sensitivity results for zones of 25, 50, 100, and 125 miles are reported below.
14
See, for example, Caves et al. (1984).
245Journal of Transport Economics and Policy Volume 35, Part 2 share of enplanements nationwide.15 Since there are four hub classi® ca- tions, a given route is in one of ten possible origin-hub-destination-hub classi® cations (for example, Large-Small). To control for density, nine dummy variables were used representing all but one of the possible combinations of service between the four airport hub classi® cations. (A Non-Hub-Non-Hub variable was not used because there were no routes in the top 1,000 in that category.) Because dummy variables were included for all hub categories in the sample, a separate constant term was not included. Slots: Four airports in the United States have limits on the number of takeoŒs and landings that may take place during any given hour. To account for the eŒect on fares of the demand restrictions at these so-called ``slot- controlled’’ airports, four dummy variables were included for the four airports: Chicago O’Hare; New York Kennedy and LaGuardia; and Washington Reagan National. The variable equalled one if the airport was on the route in question, and zero otherwise. Percentage Business: Airlines use yield management to charge business travellers higher fares than pleasure travellers. Thus, other things being equal, trip purpose is expected to aŒect airfares. Data used were the percentage of air travellers travelling between the states where the origin and destination airports were located, whose trip purpose was business.16 Concentrated Hubs: Previous research (see, for example, Borenstein, 1989) has shown that other things being equal, fares at concentrated hub airports are higher 15 Those airports that accounted for 1 per cent or more of passenger enplanements were classi® ed as Large Hubs; those enplaning 0.25 per cent or more but less than 1 per cent were classi® ed as Medium Hubs; airports enplaning 0.05 per cent or more but less than 0.25 per cent were classi® ed as Small Hubs; airports enplaning less than 0.05 per cent were classi® ed as Non Hubs. These classi® cations were determined based on the US Department of Transportation’s Airport Activity Statistics data (Form 41, Schedule T-3, Data Bank 22) for 1998. This de® nition diŒers from the US Federal Aviation Administration’s de® nition of hub class in that here airport level enplanements were 16 This was calculated from data in the US Department of Transportation’s (Bureau of Transportation Statistics), 1995 American Travel Survey. Given that state-level data were used and that the data come from an earlier year, this variable is likely to be exogenous. 246
Actual, Adjacent, and Potential Competition Morrison
than at other airports. This analysis uses the US General Accounting
O ce’s (1990) de® nition, in which an airport was considered concentrated
if it was among the top 75 in the country in terms of passenger enplane-
ments and one carrier accounted for 60 per cent or more of enplaned
passengers at the airport, or two carriers accounted for 85 per cent or more
of enplaned passengers at the airport. The resulting variable equalled one
if one or both of the airports involved on the route were concentrated
using the above de® nition.17
Regional EŒects:
To account for unobserved eŒects that are regional in nature (such as
weather), a ``Sun-Belt’’ variable was created.18 The variable equalled the
number of airports in the Sun Belt on the route in question, that is 0, 1,
or 2.
Quarterly EŒects:
To control for quarterly ® xed eŒects, three dummy variables for quarters
1± 3 were used.
Estimation Results
Full regression results are in the Appendix. Although we are interested in
the Southwest competition variables, plausible values for the other
17
This was calculated from the enplanements data in the US Department of
Transportation’ s Airport Activity Statistics data (Form 41, Schedule T-3, Data Bank
22) for 1998.
18
The de® nition of ``Sun-Belt’’ states that provided the best ® t to the data was a group
including California, Nevada, Arizona, New Mexico, Texas, Louisiana, Mississippi,
Alabama, and Florida. Alternatively, the results did not change qualitatively when
regional eŒects were captured using eight of the following nine US Department of
Transportation regional variables: Region 1 (the base case): Connecticut, Maine,
Massachusetts, New Hampshire, Rhode Island, and Vermont; Region 2: New York, New
Jersey, and Pennsylvania; Region 3: Delaware, District of Columbia, Florida, Georgia,
Maryland, North Carolina, South Carolina, Virginia, and West Virginia; Region 4:
Illinois, Indiana, Michigan, Ohio, and Wisconsin; Region 5: Alabama, Kentucky,
Mississippi, and Tennessee; Region 6: Iowa, Kansas, Minnesota, Missouri, Nebraska,
North Dakota, and South Dakota; Region 7: Arkansas, Louisiana, Oklahoma, and
Texas; Region 8: Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, and
Wyoming; Region 9: California, Oregon, and Washington.
247Journal of Transport Economics and Policy Volume 35, Part 2
coe cients will lend credence to the model. The quarterly dummies
reveal that fares are higher in quarter 1 than in the other periods (per-
haps due to a smaller fraction of price-sensitive vacation travellers in that
period). The density dummies are all statistically signi® cant, but their
relative magnitudes do not appear to show any interesting relationship.
Two of the slot coe cients have positive signs, but only one (O’ Hare) is
statistically signi® cant at the 5 per cent level. The coe cient for
Washington National is negative, but is not statistically signi® cant.
Finally, the coe cient for New York’ s Kennedy Airport has a negative
sign and is statistically signi® cant. However, slots at Kennedy are not in
eŒect throughout the entire day and, given its greater distance from New
York population centres, it may cater to more price-sensitive pleasure
travellers. The distance coe cient indicates that, as expected, fares
increase less than proportionately with distance, other things being equal,
due to the ® xed ¯ ight costs that are independent of the length of the
¯ ight (such as takeoŒ and landing). The concentrated hub dummy
indicates that fares are about 4 per cent higher on routes involving
concentrated hubs, even after controlling for route competition and other
eŒects. The percentage business coe cient indicates that fares are 28 per
cent higher on routes with 75 per cent business travellers compared with
otherwise comparable routes with 25 per cent business travellers.19 The
sun-belt variable indicates that fares involving ¯ ights to or from that
region are 6 per cent lower than elsewhere, all else being equal. The
competition variables for carriers other than Southwest seem quite rea-
sonable: additional competition lowers fares, and Low-Fare and Other
carriers have a larger impact on fares than Major carriers.20 On balance,
the model seems to produce plausible results.
Estimation results for the Southwest Airlines competition variables of
interest are reported in Table 2. All the estimated coe cients are sig-
ni® cantly diŒerent from zero at the 1 per cent level. Although all coe -
cients are precisely estimated, it is useful to see if their relative magnitudes
seem reasonable before we look at their absolute magnitude.
The relative magnitudes of the actual competition variables seem rea-
sonable because the magnitude of the coe cient declines as the service
19
Calculated as follows: exp(0.75*coe cient)/exp(0.25*coe cient)-1.
20
The diŒerential impact of Major carriers on fares compared with the relatively large
impact of Southwest and other Low-Fare carriers is consistent with the ® ndings of Bailey
et al. (1985), Strassman (1990), and Windle and Dresner (1995).
248Actual, Adjacent, and Potential Competition Morrison
Table 2
Estimation Results for Southwest Airlines Competition Variables
Category Variable Coe cient Standard Error EŒect on Fares*
Actual Competition
At-At-Serves 0:620 0:023 46:2
At-Near-Serves 0:306 0:023 26:4
Near-Near-Serves 0:167 0:029 15:4
Potential Competition
At-At-Does not Serve 0:401 0:027 33:0
At-Near-Does not Serve 0:140 0:033 13:1
Near-Near-Does not Serve 0:188 0:033 20:7
At-Far-Serves 0:125 0:021 11:8
Near-Far-Serves 0:067 0:024 6:5
Source: Author’s calculations. See Appendix for full regression results and main text for a detailed
explanation of the variables.
*Given that the regression equation explains the logarithm of fares, the eŒect on fares is exp(coe -
cient)-1.
becomes less of a substitute for travel on the route in question. The same is
true for the potential competition variables, with the exception of Near-
Near-Does not Serve.21 It is also reasonable that the magnitude of the
eŒect of actual competition is greater than the eŒect of potential compe-
tition in the two cases where a plausible eŒect of potential competition was
estimated (that is At-At-Serves vs. At-At-Does not Serve, and At-Near-
Serves vs. At-Near-Does not Serve).22 Thus, the relative magnitudes of the
coe cients seem reasonable.
As far as the absolute eŒects are concerned, when Southwest serves a
route (At-At-Serves), fares are 46 per cent lower than on otherwise com-
parable routes that it does not serve. If it serves an adjacent route (At-
Near-Serves and Near-Near-Serves), fares are from 15 to 26 per cent
below otherwise comparable routes. Potential competition from South-
21
The eŒect of Southwest in the Near-Near-Does not Serve case is to raise fares by 21 per
cent. It should be noted, however, that this estimate is based on only four quarterly
observations for one route in the sample (Dayton-Washington Reagan National).
22
Morrison and Winston (1987) estimated the impact (for all types of carriers combined) of
actual competition on a route (AAS in this paper’s terminology) and potential
competition (de® ned as carriers serving one or both airports on a route, but not the
route itself, AAX, ANX, and AFS in this paper’s terminology). They found that a
potential competitor had 1/3 the eŒect on passenger welfare that one actual competitor
had. Here, depending on the speci® c type of (comparable) potential competition
considered, the eŒect on fares of Southwest as a potential competitor ranges from 50 to
71 per cent of the eŒect of actual competition from Southwest.
249Journal of Transport Economics and Policy Volume 35, Part 2 west is most eŒective when it serves both endpoints of a route but not the route itself (lowering fares by 33 per cent) and least eŒective when it only serves one airport that is near one of the airports in question, in which case fares are reduced by 6 per cent.23 Simulation The model was used to estimate the impact that Southwest Airlines has on airfares. For each route in the complete data set of all routes (not just the top 1,000) passenger expenditure (predicted fare times number of pas- sengers) was estimated with and without Southwest Airlines.24 The results in Table 3, decomposed into the eŒect of each variable, show the savings to air travellers in 1998 because of Southwest.25 Not surprisingly, the biggest impact of competition from Southwest comes when it serves the route in question, lowering fares by nearly $7 billion. When the eŒect of competition on adjacent routes is taken into account, the total eŒect of actual competition from Southwest increases to nearly $10 billion. Adding the eŒect of Southwest as a potential competitor raises the total impact of Southwest Airlines to $12.9 billion. Given that Southwest had passenger revenue of nearly $4.0 billion in 1998, and that its presence on a route lowers fares on that route by an estimated 46.2 per cent, Southwest’s low fares were directly responsible for $3.4 billion of savings to passengers. The remaining $9.5 billion represents the eŒect that actual, adjacent, and potential competition from Southwest had on other carriers’ fares. These 23 When a comparison is possible, these results are comparable with Dresner et al. who found that fares are 41 per cent lower when Southwest serves a route and 8 to 36 per cent lower when Southwest serves an adjacent route. 24 The simulation only used those coe cients for Southwest that were statistically diŒerent from zero at the 5 per cent level of signi® cance. Because Southwest’s enplanements, passenger miles, and revenue in the sample ranged from 9.9 per cent to 10.1 per cent of their respective population values, the results of the simulation were multiplied by ten to obtain the ® gures reported in the Table. 25 These savings do not re¯ ect the change in consumer surplus but represent how much more costly the trips currently taken would be without Southwest. Assuming a constant elasticity of demand, the change in consumer surplus would be $10.9 billion with an elasticity of (minus) 0.75, $9.4 billion with an elasticity of 1.5, and $7.2 billion with an elasticity of 3.0. Furthermore, this analysis is short-run in nature and does not consider that if Southwest did not exist, other carriers may have taken its place on some of the routes served. If it was replaced on all the routes it serves, the savings attributable to Southwest would be $9.9 billion if replaced by a low-fare carrier, $11.3 billion if replaced by an ``other’’ carrier, and $12.5 billion if replaced by a major carrier. (This calculation only considers the direct eŒect of carriers other than Southwest because, as noted in footnote 11, it was not possible to estimate a plausible equation with a complete set of competition variables for all types of carriers.) 250
Actual, Adjacent, and Potential Competition Morrison
Table 3
Decomposition of Savings Attributed to
Southwest Airlines in 1998
(billions of dollars)
Category Variable EŒect
Actual Competition 9:58
At-At-Serves 6:55
At-Near-Serves 2:56
Near-Near-Serves 0:47
Potential Competition 3:30
At-At-Does not Serve 0:95
At-Near-Does not Serve 0:18
Near-Near-Does not Serve 0:01
At-Far-Serves 1:76
Near-Far-Serves 0:42
Total Savings 12:88
Source: Author’s calculations. See text for a detailed explanation of the
variables.
savings amount to 20 per cent of all domestic airline passenger revenue
and imply that Southwest is responsible for 53 per cent of the fare
reductions attributed to airline deregulation.26
Sensitivity of Results
To assess the sensitivity of the results to the radius of the assumed zone of
in¯ uence and to the inclusion of frequent ¯ ier tickets, nine additional
regressions were estimated with the radius of the zone set at 25, 50, 75,
100, and 125 miles, including and excluding frequent ¯ ier tickets. As
shown in Table 4, the estimated savings attributed to Southwest Airlines
range from $8.73 billion to $13.80 billion. Both including and excluding
frequent ¯ ier tickets, the estimated savings increase as the radius of the
zone of in¯ uenced is increased, reach their maxima at 50 miles, and then
decrease. Estimated savings including frequent ¯ ier tickets are always less
26
This is based on an update to the estimate in Morrison and Winston (1995) of the fare
reductions attributable to airline deregulation. For 1998, fares were 28 per cent lower
than an estimate of what they would have been had regulation continued. Given actual
industry domestic passenger revenue of $64 billion in 1998, this amounts to savings of
$24.4 billion. $12.9 billion is 53 percent of $24.4 billion. To the extent that the estimate of
the fare reductions from deregulation is understated, Southwest’s contribution is
overstated. But any plausible estimate of the fare reductions from deregulation still shows
Southwest making a signi® cant contribution. Morrison and Winston (1995) also found
that fare reductions represented about 2/3 of the bene® ts to passengers from airline
deregulation.
251Journal of Transport Economics and Policy Volume 35, Part 2
Table 4
Sensitivity of Savings Attributed to
Southwest Airlines in 1998
(billions of dollars)
Radius (miles) of Including Excluding
Zone of In¯ uence Frequent Flier Tickets Frequent Flier Tickets
25 9.62 10.00
50 13.19 13.80
75 12.88 13.30
100 11.00 11.45
125 8.73 9.15
Source: Author’s calculations using estimated coe cients that were signi® cant at
the 5 per cent level.
than if they are excluded. Thus, the results reported above (which were
chosen because they yielded the best ® t for regressions that included fre-
quent ¯ ier tickets) are reasonably robust.27
Comparison with Other Studies
Table 5 compares the results of this study with the results of three previous
studies of the aggregate impact of Southwest Airlines on airfares. These
previous studies assessed Southwest’s eŒect on the routes that it served,
Table 5
Comparison with Other Studies
Study Year Amount that Southwest’ s Share of
of Data Southwest Lowers Fares Domestic Scheduled
as a Percentage of Industry Revenue Passenger Miles
Domestic Passenger Revenue
Morrison and Winston (1995) 1988± 92 8.5% on routes served 2.9% (1990)
Bennett and Craun (1993) 1992 5.6% ± 6.7% on routes served 4.0% (1992)
U.S. DOT (1996) 1995 8.4% ± 10.1% on routes served 5.9% (1995)
Morrison (2000) 1998 10.2% on routes served 6.9% (1998)
4.7% on adjacent routes
5.2% from potential
competition
Sources: Listed above and Air Transport Association (various issues) and Southwest Airlines
(www.southwest.com).
27
If the simulation is limited to only those 1,000 routes used in the regression equation, the
eŒect is $9.7 billion.
252Actual, Adjacent, and Potential Competition Morrison
but did not capture the additional eŒects of competition on adjacent
routes or from potential competition.
Based on a fare model and entry and exit models, and using data from
1988-92 for the 1,000 most heavily travelled routes, Morrison and Winston
(1995) estimated that if Southwest Airlines exited the airline industry, fares
nationwide would increase by 8.5 per cent after ® ve years, even after
taking into account the subsequent entry that would probably occur.
Bennett and Craun (1993) found that if fares on Southwest’ s routes in
1992 were raised to the level of fares on non-Southwest routes, industry
revenue would rise by $2.5-$3.0 billion (holding tra c constant), which
amounts to 5.6-6.7 per cent of 1992’s domestic scheduled passenger revenue.
The US Department of Transportation (1996) estimated that low-fare
carriers saved passengers $6.3 billion during the four quarters ending in
1995:3. If we conservatively assume that Southwest’s and other low-fare
carriers’ eŒect on fares is the same, because Southwest accounted for 71
per cent of low-fare carriers’ revenue passenger miles during the four
quarters studied, Southwest’ s eŒect would be $4.5 billion, amounting to
8.4 per cent of domestic scheduled passenger revenue. If, on the other
hand, we assume that Southwest’ s presence on a route exerts a stronger
in¯ uence on fares as indicated by the estimates in this paper, Southwest’ s
eŒect rises to $5.4 billion, amounting to 10.1 per cent of domestic sched-
uled passenger revenue.28
The Table also shows Southwest’s relative size during the time periods
covered by these studies. Given that Southwest accounted for a larger
share of the industry in 1998, and that this study includes both the eŒect of
competition on adjacent routes and the eŒect of potential competition
from Southwest, which the other aggregate studies did not, the results
seem reasonable, although dramatic.29
Summary and Conclusion
This paper reports estimates of the savings that Southwest Airlines has
brought to air travellers in the United States during 1998 using an original
set of actual, adjacent, and potential competition variables. The estimated
savings due to competition from Southwest was $12.9 billion. Southwest’ s
28
Here, we assume that the eŒect of a Southwest passenger mile is 2.25 times more
powerful than other low-fare carriers’ passenger miles; 2.25 is the ratio of the eŒect of
Southwest (1-exp(-0.62)) to the eŒect of other low-fare carriers (1-exp(-0.23).
29
The US Department of Transportation (1996) notes that the impact of low-fare carriers
has been growing at an increasing rate.
253Journal of Transport Economics and Policy Volume 35, Part 2
low fares were directly responsible for $3.4 billion of these savings to
passengers. The remaining $9.5 billion represents the eŒect that actual,
adjacent, and potential competition from Southwest had on other carriers’
fares. These savings amount to 20 per cent of the airline industry’ s 1998
domestic scheduled passenger revenue and slightly more than half of the
fare reductions attributed to airline deregulation Ð a sizable impact from
a carrier that accounts for about 7 per cent of the industry’s domestic
scheduled passenger miles. From a policy perspective, these results are
both troubling and encouraging. On the one hand, it is troubling to ® nd
that a large part of the fare reductions from airline deregulation is due to
one carrier Ð albeit one, along with Paci® c Southwest Airlines, whose
performance as an intrastate carrier not regulated by the Civil Aeronautics
Board provided a natural experiment that contributed to the passage of
the Airline Deregulation Act. On the other hand, if entry by the ``right’ ’
carrier can make such a diŒerence, policies that encourage entry Ð for
example relaxing the restrictions on entry by foreign-owned carriers-may
have a large impact on passenger welfare.
References
Air Transport Association (various years): Annual Report. Washington, DC: Air Transport
Association.
Bailey, Elizabeth E., David R. Graham, and Daniel P. Kaplan (1985): Deregulating the
Airlines. Cambridge, MA: The MIT Press.
Bailey, Elizabeth E. and John C. Panzar (1981): ``The Contestability of Airline Markets
during the Transition to Deregulation.’’ Law and Contemporary Problems, 44, 125± 45.
Baumol, William J., John C. Panzar, and Robert D. Willig (1982): Contestable Markets
and the Theory of Industrial Structure. New York: Harcourt Brace Jovanovich.
Bennett, Randall D. and James M. Craun (1993): ``The Airline Deregulation Evolution
Continues: The Southwest EŒect.’’ O ce of Aviation Analysis, O ce of the Secretary,
US Department of Transportation.
Borenstein, Severin (1989): ``Hubs and High Fares: Dominance and Market Power in the
US Airline Industry.’’ Rand Journal of Economics, 20, 344± 65.
Breyer, Stephen (1989): Regulation and its Reform. Cambridge: Harvard University Press.
Caves, Douglas W., Laurits R. Christensen, and Michael W. Tretheway (1984):
``Economies of Density versus Economies of Scale: Why Trunk and Local Service
Airlines Costs DiŒer.’’ Rand Journal of Economics, 15, 471± 89.
Dresner, Martin, Jiun-Sheng Chris Lin, and Robert Windle (1996): ``The Impact of Low-
Cost Carriers on Airport and Route Competition.’’ Journal of Transport Economics and
Policy, 30, 309± 28.
Morrison, Steven A. and CliŒord Winston (1987): ``Empirical Implications and Tests of
the Contestability Hypothesis.’’ Journal of Law and Economics, 30, 53± 66.
Morrison, Steven A. and CliŒord Winston (1995): The Evolution of the Airline Industry.
254Actual, Adjacent, and Potential Competition Morrison
Washington, DC: The Brookings Institution.
Strassmann, Diana L. (1990): ``Potential Competition in the Deregulated Airlines.’’ The
Review of Economics and Statistics, 72, 696± 702.
Transportation Research Board, National Research Council (1999): Entry and Competition
in the US Airline Industry: Issues and Opportunities. Washington, DC: National
Academy Press.
US General Accounting O ce (1990): Airline Competition: Higher Fares and Reduced
Competition at Concentrated Airports. GAO/RCED-90-102, Washington, DC.
US Department of Transportation (1996): ``The Low Cost Airline Service Revolution.’’
Windle, Robert and Martin Dresner (1995): ``The Short and Long Run EŒects of Entry on
US Domestic Air Routes.’’ Transportation Journal, 35, 14± 25.
255Journal of Transport Economics and Policy Volume 35, Part 2
Appendix
Ordinary Least Squares Regression Results
Dependent Variable: Logarithm of Average Fare
White Heteroskedasticity-Consistent Standard Errors
Variable Coe cient Standard Error t-Statistic Probability
Quarter 1 Dummy 0.052490 0.01053 9 4.98075 3 0.0%
Quarter 2 Dummy 0.015028 0.00977 2 1.53786 7 12.4%
Quarter 3 Dummy 0.000123 0.00983 1 0.01249 3 99.0%
Large Hub-Large Hub Dummy 2.377770 0.05973 5 39.80539 0 0.0%
Large Hub-Medium Hub Dummy 2.364472 0.05876 3 40.23772 0 0.0%
Large Hub-Small Hub Dummy 2.268900 0.05629 8 40.30147 0 0.0%
Large Hub-Non Hub Dummy 2.053422 0.06593 4 31.14338 0 0.0%
Medium Hub-Medium Hub Dummy 2.305377 0.05921 5 38.93204 0 0.0%
Medium Hub-Small Hub Dummy 2.247348 0.06119 1 36.72657 0 0.0%
Medium Hub-Non Hub Dummy 2.217791 0.06889 0 32.19310 0 0.0%
Small Hub-Small Hub Dummy 1.956237 0.05487 5 35.64899 0 0.0%
Small Hub-Non Hub Dummy 2.177449 0.08914 7 24.42544 0 0.0%
Washington Reagan National Dummy 0.022675 0.021961 1.03251 7 30.2%
New York Kennedy Dummy 0.159651 0.026972 5.91912 9 0.0%
New York LaGuardia Dummy 0.029887 0.01591 5 1.87797 3 6.1%
Chicago O’Hare Dummy 0.068581 0.01812 3 3.78423 0 0.0%
Percent Business Travel 0.494438 0.024715 20.00529 0 0.0%
Distance (logarithm) 0.408168 0.007605 53.67442 0 0.0%
Number of Airport in Sunbelt 0.057146 0.006638 8.60893 8 0.0%
Concentrated Hub Dummy 0.055875 0.009511 5.87500 8 0.0%
Southwest: At-At-Serves 0.619880 0.022826 27.15637 0 0.0%
Southwest: At-Near-Serves 0.306484 0.023236 -13.19014 0 0.0%
Southwest: Near-Near-Serves 0.167483 0.028573 5.86166 1 0.0%
Southwest: At-At-Does Not Serve 0.401170 0.027317 14.68575 0 0.0%
Southwest: At-Near-Does Not Serve 0.140334 0.033032 4.24836 4 0.0%
Southwest: Near-Near-Does not Serve 0.188314 0.032628 5.77155 4 0.0%
Southwest: At-Far-Serves 0.124500 0.020706 6.01268 9 0.0%
Southwest: Near-Far-Serves 0.066610 0.023775 2.80165 7 0.5%
Number of Low Fare Carriers on Route 0.233511 0.010366 22.52553 0 0.0%
Number of Major Carriers on Route 0.029259 0.004696 6.23018 4 0.0%
Number of Other Carriers on Route 0.117270 0.020069 5.84335 4 0.0%
Number of Observations = 4,000
R2 ˆ 0:739
Source: Author’s calculation. See text for a detailed explanation of the variables.
256You can also read
